Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
PIZZA.BAT
Nov 12, 2016





Yeah as long as the bubble of air doesnít settle in the pump or heat sink youíre fine. AIOs ship with a small pocket of air for a reason

Adbot
ADBOT LOVES YOU

Cloner of the Elks
Jul 2, 2006







My first ever attempt at this.

1x Black 320mm EK Radiator (top)
1x White 320mm Corsair XR7 Radiator (bottom)
1x White Corsair XD5 reservoir/pump combo
1x White Corsair XC7 CPU water block
1x Bykski RX580 water block (lol)
19x White Barrow fittings
Mayhems X1 Eco Blood Red liquid

Amazing photo with the glass still on the case (which is the Phanteks knock-off version of the O11 Dynamic - has a white panel with LED strip).


There are less bubbles in the reservoir now, but they still going away slowly. Haven't got a pic of the pump outlet with drain valve, but it is there.
The Corsair logo LED on the CPU block and reservoir are set to change color with temps.
I'll be getting a set of cable crimps and sleeves around Christmas time so going to try my hand at replacing those awful graphics card wires with something nicer looking.

jink
May 8, 2002

Drop it like it's Hot.

Taco Defender

Cloner of the Elks posted:

My first ever attempt at this.


Holy. poo poo.

Fantastic first ever attempt!

Cloner of the Elks
Jul 2, 2006







jink posted:

Holy. poo poo.

Fantastic first ever attempt!

Thank you!
The hardest tube was going from the CPU to top of reservoir. Was also a pain going from pump to the white rad at the bottom.

Will have to try and get around to getting photos of that, too

Collateral Damage
Jun 13, 2009



Brave going hardline for your first build. Good job!

jink
May 8, 2002

Drop it like it's Hot.

Taco Defender

Collateral Damage posted:

Brave going hardline for your first build. Good job!

Yeah. This.

I am a big pussy and went soft tubing. It still sucked.
I canít imagine hardline to start!

That being said, I want to go hardline. It is far more resilient to decay over time and takes up less space...but I need to buy fittings again. I could have bought a used car instead of these Bitspower purchases. 😓

ilkhan
Oct 7, 2004

IF I JUST LICK ENOUGH BOOT LEATHER, BIG DADDY TRUMP WILL SURELY LOVE ME

jink posted:

Yeah. This.

I am a big pussy and went soft tubing. It still sucked.
I can’t imagine hardline to start!

That being said, I want to go hardline. It is far more resilient to decay over time and takes up less space...but I need to buy fittings again. I could have bought a used car instead of these Bitspower purchases. 😓
I'll just quote this for me too. Probably when I add a GPU block to my loop. 3x120+2x140 for just a 65w 3600 cpu... I leave the 120s off cause it ain't like I need them spinning and they are slightly noisier than the 140s.

ilkhan fucked around with this message at 22:57 on Dec 24, 2020

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

Meanwhile I'm gonna see just how toasty a system can get with a 5600X + 3080 on 2x120 rads in a case known for lovely airflow! It'll be fun. And warm. And might end up with me hanging a 240 rad out the back or something equally silly.

Sagebrush
Feb 26, 2012


"Why does that Subaru break down every time you look at it, Travis", Punchy said. I nearly fell out of the jump seat in my Brat, aghast. "That thing a princess?" I coughed and gulped. "Hahahaha, nice one, Punchy", I said

DrDork posted:

I was under the impression that air is effectively incompressible until you get it moving reasonably quick (100m/s IIRC?)

You may have this impression because in fluid dynamics air is usually treated as if it's incompressible when it's a low-energy (low pressure or velocity) situation. In the context of a light airplane, for instance, the pressure gradients are quite small -- a Cessna 172 in level flight has about 0.1 psi differential between the lower and upper wing surfaces -- and the airplane's velocity through the air is low enough that pressure changes across the airframe can be considered to be instantaneous.

When you approach supersonic speeds the air can't get out of its own way any more, and you do have to consider compressibility.

skylined!
Apr 6, 2012

THE DEM DEFENDER HAS LOGGED ON


So I reconfigured my loop this past week, just finishing tonight, because I got a 3080 FE to swap into it. I tossed the FE in the case on air while waiting for a waterblock and segregated the loop to the processor only. Idle temps dropped to almost ambient (about 21C) and under load didn't really get above 50C. CPU is a 9900k at stock currently.

Since adding the GPU back in the loop, idle temps are around 45C on the CPU and 30C on the GPU in a 19C room. Can't really figure out why; does the 3080 really add that much heat energy to a loop?!

I am running two of this big bastards with a G1 distro plate for the o11 XL and included pump. My fans aren't great (LL120s) and several aren't working because my commander pro is a piece of poo poo and waiting on an RMA, but.. still. This is a wild temp swing. Anyone have any ideas?

Edit so right after posting this an air bubble pushed through and idle temps on the CPU dropped to like 24-34C lol, man this poo poo is so unnecessarily stressful I am going to bed.

Here have a picture for reading this dumb post. Working on custom cables tomorrow then closing back up.

.

skylined! fucked around with this message at 04:34 on Dec 30, 2020

Indiana_Krom
Jun 18, 2007
Net Slacker

skylined! posted:

So I reconfigured my loop this past week, just finishing tonight, because I got a 3080 FE to swap into it. I tossed the FE in the case on air while waiting for a waterblock and segregated the loop to the processor only. Idle temps dropped to almost ambient (about 21C) and under load didn't really get above 50C. CPU is a 9900k at stock currently.

Since adding the GPU back in the loop, idle temps are around 45C on the CPU and 30C on the GPU in a 19C room. Can't really figure out why; does the 3080 really add that much heat energy to a loop?!

I am running two of this big bastards with a G1 distro plate for the o11 XL and included pump. My fans aren't great (LL120s) and several aren't working because my commander pro is a piece of poo poo and waiting on an RMA, but.. still. This is a wild temp swing. Anyone have any ideas?

Edit so right after posting this an air bubble pushed through and idle temps on the CPU dropped to like 24-34C lol, man this poo poo is so unnecessarily stressful I am going to bed.

Here have a picture for reading this dumb post. Working on custom cables tomorrow then closing back up.

.

Air bubbles may indicate the flow rate is a bit on the low side. Perhaps with that distribution plate, two radiators, and a couple blocks your pump is starting to bog. Although the flow rate doesn't *have* to be super fast.

skylined!
Apr 6, 2012

THE DEM DEFENDER HAS LOGGED ON


Indiana_Krom posted:

Air bubbles may indicate the flow rate is a bit on the low side. Perhaps with that distribution plate, two radiators, and a couple blocks your pump is starting to bog. Although the flow rate doesn't *have* to be super fast.

I was thinking the same but after about two hours of running (after force-pegging the pump to 100% in bios) it seems everything is flowing much better. I also had to pick up the whole thing and rotate it a bunch to shake bubbles loose in the radiators and blocks. But yea, keeping a close eye on this pump...

jink
May 8, 2002

Drop it like it's Hot.

Taco Defender

skylined! posted:

So I reconfigured my loop this past week, just finishing tonight, because I got a 3080 FE to swap into it. I tossed the FE in the case on air while waiting for a waterblock and segregated the loop to the processor only. Idle temps dropped to almost ambient (about 21C) and under load didn't really get above 50C. CPU is a 9900k at stock currently.

Since adding the GPU back in the loop, idle temps are around 45C on the CPU and 30C on the GPU in a 19C room. Can't really figure out why; does the 3080 really add that much heat energy to a loop?!

I am running two of this big bastards with a G1 distro plate for the o11 XL and included pump. My fans aren't great (LL120s) and several aren't working because my commander pro is a piece of poo poo and waiting on an RMA, but.. still. This is a wild temp swing. Anyone have any ideas?

Edit so right after posting this an air bubble pushed through and idle temps on the CPU dropped to like 24-34C lol, man this poo poo is so unnecessarily stressful I am going to bed.

Here have a picture for reading this dumb post. Working on custom cables tomorrow then closing back up.

.

Jeez. That looks fantastic. I can't comment on the temps but just wanted to say that I admire your build.

The freaking 3080FE block is SO AWESOME! Small, great looks, back distro block... oh baby.

skylined!
Apr 6, 2012

THE DEM DEFENDER HAS LOGGED ON


jink posted:

Jeez. That looks fantastic. I can't comment on the temps but just wanted to say that I admire your build.

The freaking 3080FE block is SO AWESOME! Small, great looks, back distro block... oh baby.

Ya the FE blocks in general look great; I am enjoying the port placement on these blocks. I like the EK block a lot too but didn't want to wait and also pay twice as much for it (this is the bitspower block in the picture).

ilkhan
Oct 7, 2004

IF I JUST LICK ENOUGH BOOT LEATHER, BIG DADDY TRUMP WILL SURELY LOVE ME

EK announced a 3070 block. So we'll probably see more of those. Glad to see it. Wonder if they'll also fit 3060ti cards...

jink
May 8, 2002

Drop it like it's Hot.

Taco Defender

skylined! posted:

Ya the FE blocks in general look great; I am enjoying the port placement on these blocks. I like the EK block a lot too but didn't want to wait and also pay twice as much for it (this is the bitspower block in the picture).

For real, that's killer.

I was unable to find any 3080 for sale (like everyone else), and tried to get a 3070. That wasn't going well for me so I landed on a 2080ti for same price... and... IT IS A MASSIVE BEAST THAT I CANNOT EVEN FIT IN MY CASE due to the waterblock being so drat big

The 3080 block setup is so nice. I am jelly. :P

skylined!
Apr 6, 2012

THE DEM DEFENDER HAS LOGGED ON


jink posted:

For real, that's killer.

I was unable to find any 3080 for sale (like everyone else), and tried to get a 3070. That wasn't going well for me so I landed on a 2080ti for same price... and... IT IS A MASSIVE BEAST THAT I CANNOT EVEN FIT IN MY CASE due to the waterblock being so drat big

The 3080 block setup is so nice. I am jelly. :P

I basically gave up until I got a notification for Best Buy from a chrome extension I forgot I installed. Because of how Best Buy handles sales (seems to be regionally) I was able to get the FE before it sold out. Bought on Tuesday and picked up Sunday! I think the extension was popcart or something.

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

PSA: today I hooked up my EK 3080 FE super sweet looking waterblock only to have it leak all over the place from the frame. Upon further investigation it looks like several of the bolts that hold the metal plate to the acrylic window were not tightened down sufficiently. So if you end up getting your hands on one, I highly recommend giving them all a good twist before you do anything else with it.

jink
May 8, 2002

Drop it like it's Hot.

Taco Defender

skylined! posted:

I basically gave up until I got a notification for Best Buy from a chrome extension I forgot I installed. Because of how Best Buy handles sales (seems to be regionally) I was able to get the FE before it sold out. Bought on Tuesday and picked up Sunday! I think the extension was popcart or something.

drat that is incredible.

I 'settled' for a 3080 I found from a redditor with a Bitspower block preinstalled. Works great.

I lust after a 3080FE but... I have a 3080 and I should just be happy.


DrDork posted:

PSA: today I hooked up my EK 3080 FE super sweet looking waterblock only to have it leak all over the place from the frame. Upon further investigation it looks like several of the bolts that hold the metal plate to the acrylic window were not tightened down sufficiently. So if you end up getting your hands on one, I highly recommend giving them all a good twist before you do anything else with it.

Wow. That's a terrible feeling.

I've cracked Acrylic too many times and the -drip drip- sucks!

SPACE HOMOS
Jan 12, 2005



DrDork posted:

PSA: today I hooked up my EK 3080 FE super sweet looking waterblock only to have it leak all over the place from the frame. Upon further investigation it looks like several of the bolts that hold the metal plate to the acrylic window were not tightened down sufficiently. So if you end up getting your hands on one, I highly recommend giving them all a good twist before you do anything else with it.

You should probably contact EKWB about that. The blocks are supposed to be QC'd and leak tested. I wish the areas where fittings go wasn't acrylic on all the new rgb stuff.

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

SPACE HOMOS posted:

You should probably contact EKWB about that. The blocks are supposed to be QC'd and leak tested. I wish the areas where fittings go wasn't acrylic on all the new rgb stuff.

After further investigation I located the actual screws that keep the thing together. About half of them were loose enough I could turn them 1/2 or more of a turn by hand without much effort. So either they weren't tested prior to shipping, or the shipping between Slovenia and the US was rough enough to vibrate them out a bit (doesn't look like they had any locktite or similar on them).

Broke the entire thing down and found that the main O-ring was miss-seated and pinched--not something rough shipping is likely to have caused. Reseated it and now it's working fine. Honestly, while it looks loving amazing, I'm not terribly impressed with the manufacturing quality. I'll admit this is my first aftermarket full waterblock, but I expected it to be...smoother. There are CNC marks all over the plate, including the section that contacts the die--from other watercooling products I'm used to those areas being pretty smooth, if not actually lapped (like I'd expect from a $300 premium product). The visible side is all nice and smoothed, why not smooth out the side were smoothness actually matters? The edges along the cut portions were also often rough enough to snag the cloth and cotton balls I was using to clean and prep it, moreso than I would have expected.

Basically it seems like the block itself needed a polishing pass that it didn't get. Unsure if that's just how GPU waterblocks are and everyone accepts it, but between that and the obvious QA miss in terms of leakage make it seem like I got a rushed product.

On the upside, now that it's installed it looks baller as hell, and it turns out that 2x120mm rads are mostly enough to handle a 3080+5600X, though they're struggling under 100% load because they're an an Evolv Shift (glass) which has the thermal characteristics of your average kitchen over, and I've got some UV light strips in there which are surprisingly hot.

Riflen
Mar 12, 2009

"Cheating bitch"

Bleak Gremlin

This isn't how it is in general at all, but EK are not regarded as a quality brand. They farm out manufacture to China. If you're wanting the best quality blocks for your money and less interested in how speedily they're released or flashy they look, then I would recommend WaterCool and AquaComputer.
I've had 8 GPU blocks from these companies over the years and never once was there a single problem with any of them. I'm sure I'll be using them next time too.

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

Can't say I've heard too many people say EK is a bad brand before.

Not that it really matters too much--there just aren't many other options. Yeah, AquaComputer has a 3080 block in theory, you just can't buy it. Corsair has one, but it's both ugly as sin and they've had some pretty bad QA issues in the past with their waterblocks. BitsPower has one, too, but they're literally a Chinese company and even more expensive. Aaaaand that's about it.

Riflen
Mar 12, 2009

"Cheating bitch"

Bleak Gremlin

DrDork posted:

Can't say I've heard too many people say EK is a bad brand before.

Not that it really matters too much--there just aren't many other options. Yeah, AquaComputer has a 3080 block in theory, you just can't buy it. Corsair has one, but it's both ugly as sin and they've had some pretty bad QA issues in the past with their waterblocks. BitsPower has one, too, but they're literally a Chinese company and even more expensive. Aaaaand that's about it.

And Bykski is another Chinese company that rips off EK designs most of the time. I've learned to shop for the GPU based on the block I want to use.

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

This is the first time I've bothered with a GPU waterblock, because up until now they made absolutely no sense outside an aesthetic option: previous gens you could just get a G12 and a $75 AIO and slap that bad boy on there and be good, and be able to port that over to almost any other card. Made it a whole lot harder to defend spending $200+ on a full block cooler that is locked to that card and that card alone, even if you already have a CLC for your CPU.

(though I will readily admit "I was ok spending the cash because it made it look cooler" is a perfectly valid reason for getting one)

Buuuut thanks to Ampere changing the mounting pattern, that's no longer possible (at least until NZXT bothers to update it), and as said, there aren't many (any) options for 3080 blocks that aren't from EK or someone even lower on the food chain, so here we are.

Canna Happy
Jul 11, 2004
The engine, code A855, has a cast iron closed deck block and split crankcase. It uses an 8.1:1 compression ratio with Mahle cast eutectic aluminum alloy pistons, forged connecting rods with cracked caps and threaded-in 9 mm rod bolts, and a cast high

DrDork posted:

Broke the entire thing down and found that the main O-ring was miss-seated and pinched--not something rough shipping is likely to have caused. Reseated it and now it's working fine. Honestly, while it looks loving amazing, I'm not terribly impressed with the manufacturing quality. I'll admit this is my first aftermarket full waterblock, but I expected it to be...smoother. There are CNC marks all over the plate, including the section that contacts the die--from other watercooling products I'm used to those areas being pretty smooth, if not actually lapped (like I'd expect from a $300 premium product).

O-ring problems, just like I figured. Does your block look like https://www.reddit.com/r/watercooling/comments/jubf51/ek_3080_vector_block_just_came_in_today_poor/ that? Also, did it have an ek sticker on the block saying it passed qc? In that past, all ek blocks I've purchased had a sticker saying they were pressure checked or whatever. If it looks like that reddit post, I would for sure contact customer support. Sadly, ek is cashing in on their large brand recognition with insane prices and poor quality these days.

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

Canna Happy posted:

O-ring problems, just like I figured. Does your block look like https://www.reddit.com/r/watercooling/comments/jubf51/ek_3080_vector_block_just_came_in_today_poor/ that? Also, did it have an ek sticker on the block saying it passed qc? In that past, all ek blocks I've purchased had a sticker saying they were pressure checked or whatever. If it looks like that reddit post, I would for sure contact customer support. Sadly, ek is cashing in on their large brand recognition with insane prices and poor quality these days.

Not quite that bad everywhere--if you look at the depressed areas with the more uniform machining marks, it looks more like that, rather than the die area there which looks like it got shot-peened or something. I could understand it on the VRM/VRAM lands because you're just using thermal pads there, so having it be perfectly smooth isn't terribly important, but you'd think they'd give the die portion a quick lap at least.

The actual performance seems fine, though--at 100% load it's keeping the die at 75C and the VRAM cool enough that there's no throttling like on the stock HSF (which would hit 91C and VRAM throttle, depending on which part was being stressed). It'd probably run cooler than that if I wasn't putting a 5600X into the loop as well and then trying to cool it all with 2x120mm rads in a terrible airflow case.

I couldn't find any sticker of any kind on the block, FWIW.

e; since it's a high-demand, back-ordered part, I could see someone at EK deciding to just ship them direct instead of receiving them, testing them, repacking them, etc., to get them in people's hands sooner, with the assumption that only a small percentage would have issues in the first place so it'd be better for 95% of their customers. Obnoxious to be one of the 5%, though!

e2; doing some fan tuning also has revealed that the Arctic P14 I have as the downward facing exhaust rattles like hell when pushed above 90%. So that's annoying; I might have to see about either slightly sanding down whichever of the fins is causing it, or see if I have a spare 140mm Noctura hanging out in my closet somewhere.

DrDork fucked around with this message at 23:04 on Jan 18, 2021

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

To EK's credit, they got back to my ticket the next business day, were very apologetic, and have offered to send me a replacement o-ring and a bottle of replacement coolant to make up for what I used diagnosing the issue (or a RMA, if I was cool dealing without a system for weeks).

Obviously still annoying the issue existed in the first place, but at least they didn't try to argue with me or insist I must have hosed it up myself or something.

Ak Gara
Jul 29, 2005

That's just the way he rolls.

Jesus christ lining the inside of my tower with black acrylic might have been a bad idea. What's a good anti static wipe that actually repels dust, but doesn't use alcohol? All I can find on amazon are essentially normal wet wipes.

Warmachine
Jan 30, 2012

I came here to laugh at you





Ak Gara posted:

Jesus christ lining the inside of my tower with black acrylic might have been a bad idea. What's a good anti static wipe that actually repels dust, but doesn't use alcohol? All I can find on amazon are essentially normal wet wipes.

Clearcoat it with a matte or gloss lacquer and just use normal anti-static wipes.

Ak Gara
Jul 29, 2005

That's just the way he rolls.

I think I'm gonna have to do something, even when it's all finished it's going to be a dust magnet.

Finally got around to putting in the motherboard, M.2 drives, RAM, CPU, and CPU block. One of the two reservoirs too. Still figuring out the tube pathing.
I'm thinking from left reservoir to GPU, to left lower radiator, then straight up to top radiator, to CPU, from CPU to right hand side reservoir. Or I might do dual loop, or a figure 8. Options!





DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

Just don't be stupid and do what I did, which was not think about things and end up going Res -> Rad -> Rad -> GPU -> CPU -> Res. Turns out if you do that, the GPU stays reasonably cool, but the CPU freaks the gently caress out because you're feeding it like 75C water. Whoops.

Switching CPU and GPU ordering there bumped my GPU temp by 1-2C under full load, and dropped my CPU temp by like 10-15C. Now the system doesn't feel like it always needs to run the fans on high!

ilkhan
Oct 7, 2004

IF I JUST LICK ENOUGH BOOT LEATHER, BIG DADDY TRUMP WILL SURELY LOVE ME

You should not be seeing wild water temp differences like that. Lots of testing says the water should be within 1-2c difference anywhere in the loop. How slow is your pump running? You should also be running the fans based on a water temp sensor.

Indiana_Krom
Jun 18, 2007
Net Slacker

DrDork posted:

Just don't be stupid and do what I did, which was not think about things and end up going Res -> Rad -> Rad -> GPU -> CPU -> Res. Turns out if you do that, the GPU stays reasonably cool, but the CPU freaks the gently caress out because you're feeding it like 75C water. Whoops.

Switching CPU and GPU ordering there bumped my GPU temp by 1-2C under full load, and dropped my CPU temp by like 10-15C. Now the system doesn't feel like it always needs to run the fans on high!

That is highly unlikely unless your pump was impossibly slow or just stalled out completely. Loop order should be entirely irrelevant assuming your pump is operating normally because if your pump is anything like mine it circulates the entire volume of coolant in seconds at most. The temperature across the whole loop should be within fractions of a degree at any point you measure it. In fact it should take real effort and carefully calibrated tools to even be able to measure the temperature differences in the coolant between inlet and exit points of blocks/radiators/etc. (The D5 pump I have is rated for 1500l/hr flow rate, assume its only managing half that after restrictions so 750l/hr, well the entire fluid volume of my loop is about 850 ml, so 3600 seconds divided by 750l multiplied by .85l volume equals 4.08 seconds to circulate the entire loop.)

I don't think even accidentally plumbing a block backwards would have that kind of impact (performance will be worse is the flow is reversed through a jet plate, but not *that* much worse). Perhaps what happened is you had bad contact/paste on your CPU heat spreader and when you redid everything in the different order it got tightened up and the paste started working properly. In order to raise the temperature of water to 75C in the space of just one water block, it would have to be hundreds of degrees.

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

ilkhan posted:

You should not be seeing wild water temp differences like that. Lots of testing says the water should be within 1-2c difference anywhere in the loop. How slow is your pump running? You should also be running the fans based on a water temp sensor.

Well I'm not really running a "recommended" loop here in the first place: I've got a 3080 + 5600X cooled by just a pair of 120mm rads. The pump is this Chinese special integrated CPU waterblock / pump / "res" deal, and runs at ~4000RPM constantly since apparently putting in a variable speed pump was within their budget, but allowing that pump speed to be controlled intelligently was not, so instead they just have a rheostat on a pair of wires to control the speed. And if that wasn't bad enough, I shoved it all in a Evolv shift. Yes, the type with the glass panels. Oh, and I don't have a water-temp sensor, but in retrospect that might not be the worst thing given the temp differentials.

So, yeah. Not a whole lot of rad space, hilariously poor case airflow, a "res" that's like...10ml? There's basically no spare fluid beyond what's in the tubes/blocks/rads. Frankly I'm kinda surprised this thing works at all. It's a monument to "you were so preoccupied with whether you could, you didn't stop to think if you should." (but it looks baller and it works so gently caress it)

Doing it this way means that the water gets pretty hot, which means that those rads actually do a lot of good work, because the higher the delta between your water and ambient, the more heat they can dump out. So I'd bet that the delta-T of the water before vs after the two rads is considerably higher than what it is in a more properly specced setup. The rads are non-directional, FWIW, as is the GPU block, so there was no opportunity to screw that up. The only thing that's directional at all is the pump, naturally.

But no joke, I've been running a buttcoin miner to help test the thermal abilities of this thing (because I can control how much GPU wattage it's using, anywhere between 200-320W, and because I can run it for long enough to heat-soak everything fully while still dicking around doing other stuff), and going GPU -> CPU had that poor Zen3 around 80-85 sustained, and 90C+ if it actually needed to clock up and do anything CPU-wise. Now that it's CPU -> GPU it's closer to 70C sustained, with no other changes--no need to remove the block from the CPU, since all I needed to do was swap the input and output feeds, which had enough slack in the lines already that nothing needed to be removed/replaced, just unplugged from one and moved over about 3cm.

I'm sure 75C water was an exaggeration. It's probably closer to 55C, since the GPU loads at 65C and the CPU at 75C. But yes, shoving 300W of heat output immediately before the CPU was very definitely causing higher CPU temps. If you run a more "sane" configuration where the water never heats up all that much in the first place, it probably doesn't make all that much of a difference. But with this thing, there absolutely are hotter/colder portions of the loop.

DrDork fucked around with this message at 01:25 on Jan 25, 2021

Warmachine
Jan 30, 2012

I came here to laugh at you





DrDork posted:

Well I'm not really running a "recommended" loop here in the first place: I've got a 3080 + 5600X cooled by just a pair of 120mm rads. The pump is this Chinese special integrated CPU waterblock / pump / "res" deal, and runs at ~4000RPM constantly since apparently putting in a variable speed pump was within their budget, but allowing that pump speed to be controlled intelligently was not, so instead they just have a rheostat on a pair of wires to control the speed. And if that wasn't bad enough, I shoved it all in a Evolv shift. Yes, the type with the glass panels. Oh, and I don't have a water-temp sensor, but in retrospect that might not be the worst thing given the temp differentials.

So, yeah. Not a whole lot of rad space, hilariously poor case airflow, a "res" that's like...10ml? There's basically no spare fluid beyond what's in the tubes/blocks/rads. Frankly I'm kinda surprised this thing works at all. It's a monument to "you were so preoccupied with whether you could, you didn't stop to think if you should." (but it looks baller and it works so gently caress it)

Doing it this way means that the water gets pretty hot, which means that those rads actually do a lot of good work, because the higher the delta between your water and ambient, the more heat they can dump out. So I'd bet that the delta-T of the water before vs after the two rads is considerably higher than what it is in a more properly specced setup. The rads are non-directional, FWIW, as is the GPU block, so there was no opportunity to screw that up. The only thing that's directional at all is the pump, naturally.

But no joke, I've been running a buttcoin miner to help test the thermal abilities of this thing (because I can control how much GPU wattage it's using, anywhere between 200-320W, and because I can run it for long enough to heat-soak everything fully while still dicking around doing other stuff), and going GPU -> CPU had that poor Zen3 around 80-85 sustained, and 90C+ if it actually needed to clock up and do anything CPU-wise. Now that it's CPU -> GPU it's closer to 70C sustained, with no other changes--no need to remove the block from the CPU, since all I needed to do was swap the input and output feeds, which had enough slack in the lines already that nothing needed to be removed/replaced, just unplugged from one and moved over about 3cm.

I'm sure 75C water was an exaggeration. It's probably closer to 55C, since the GPU loads at 65C and the CPU at 75C. But yes, shoving 300W of heat output immediately before the CPU was very definitely causing higher CPU temps. If you run a more "sane" configuration where the water never heats up all that much in the first place, it probably doesn't make all that much of a difference. But with this thing, there absolutely are hotter/colder portions of the loop.

Yeah, none of these numbers make sense to me. Are you actually running a coolant temperature sensor, or are you reading core temps out of HWinfo?

edit: except the part about you running a 3080 and 5600x on 240mm of radiator volume. I'm not surprised at all about high temperatures in general.

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

As noted, I don't have a coolant thermometer. The GPU temp I'm getting from Afterburner and the CPU temp from Ryzen Master. Those are, as far as things go, pretty authoritative sources for what those bits of hardware are running at. Even if the absolute numbers are somehow off, the fact that they changed after swapping the tubes to re-order things (while doing literally nothing else) remains valid.

Water temp I'm guessing at judging by how hot it still was ~10 min after turning it off so I could rearrange the tubes, at which point I'd ballpark it at 45C or so. Considering no pumps were running during that time, an operating load temp in the low 50C range seems perfectly reasonable.

Warmachine
Jan 30, 2012

I came here to laugh at you





DrDork posted:

As noted, I don't have a coolant thermometer. The GPU temp I'm getting from Afterburner and the CPU temp from Ryzen Master. Those are, as far as things go, pretty authoritative sources for what those bits of hardware are running at. Even if the absolute numbers are somehow off, the fact that they changed after swapping the tubes to re-order things (while doing literally nothing else) remains valid.

Water temp I'm guessing at judging by how hot it still was ~10 min after turning it off so I could rearrange the tubes, at which point I'd ballpark it at 45C or so. Considering no pumps were running during that time, an operating load temp in the low 50C range seems perfectly reasonable.

Do me a favor and do an OCCT power virus test on the CPU, GPU and CPU+GPU with your fans maxed at 100%, for 20 minutes each (or until TJmax is reached) with the max temps for CPU and GPU in each test recorded. Something still seems screwy here.

Indiana_Krom
Jun 18, 2007
Net Slacker

Warmachine posted:

Yeah, none of these numbers make sense to me. Are you actually running a coolant temperature sensor, or are you reading core temps out of HWinfo?

edit: except the part about you running a 3080 and 5600x on 240mm of radiator volume. I'm not surprised at all about high temperatures in general.
I read it that way the first time too, but he said it is a pair of 240MM radiators not a single one. Can't really say if it is not doing well or not because I don't have a 3080 to compare with.

Also a quick and dirty way to guess the coolant temp is to heat it up with a full stress load, then switch to the lowest power idle you can get. Wait 10-15 seconds, whatever temperature the no load components settle on is probably about 2-3C over whatever the coolant temperature is. The coolant will always be colder than whatever is the coldest chip in the loop. So if your GPU is before the CPU in the loop and you run furmark on it with no load on the CPU, and the CPU idles at 31C, then your coolant coming out of the GPU is guaranteed less than 31C.

Indiana_Krom fucked around with this message at 02:29 on Jan 25, 2021

Adbot
ADBOT LOVES YOU

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

Indiana_Krom posted:

I read it that way the first time too, but he said it is a pair of 240MM radiators not a single one. Can't really say if it is not doing well or not because I don't have a 3080 to compare with.

Also a quick and dirty way to guess the coolant temp is to heat it up with a full stress load, then switch to the lowest power idle you can get. Wait 10-15 seconds, whatever temperature the no load components settle on is probably about 2-3C over whatever the coolant temperature is. The coolant will always be colder than whatever is the coldest chip in the loop. So if your GPU is before the CPU in the loop and you run furmark on it with no load on the CPU, and the CPU idles at 31C, then your coolant coming out of the GPU is guaranteed less than 31C.

Nope, I have a pair of 120mm rads. Specifically 120mm x 28mm, so they're on the thin side (otherwise only one fits). Behold my house of poor choices!

As far as "doing well," keeping a 320W GPU under 70C in a lovely case is already asking a fair amount from 240mm worth of rads. Throwing another 50-75W of CPU in there with a X570 chipset and some lights belching excess heat (why the hell are the UV lights that hot?) and I'm actually very pleasantly surprised.

I'll run the tests here in a few and get back to you two, but I'm not sure what you're looking for out of them: I'm pumping 300-400W of heat into a system with no effective res in a case well known for lovely airflow, and while 2x120 rads are keeping things below thermal limiting break points, they're obviously undersized for the task at hand (at least in the sense that everyone going into watercooling seems to obsess over keeping temps ultra low--this setup actually ends up being considerably quieter and cooler than the air setup beforehand, and they're no longer thermally limited, so in that sense the rads could be argued to be viably, if not optimally, sized for this).

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply